🦀 In Rust We Trust — Everyone Else Gets Audited by Claude
Full-Stack Engineering · DevOps & Self-Hosting · Cybersecurity · Local AI · Open Source
Building systems under the Fuji moon, with Shiori-chan keeping my place in every story. 🦊🌸
name: Jae Jin
handle: '@Kadantte'
location: Republic of Korea 🇰🇷
pronouns: He/Him
role: Founder & Developer @ Oppa TECH
focus: Full-stack · DevOps · Cybersecurity · AI/ML
interests: Anime · Manga · Games · Graphic Design
mission: Build it well. Secure it. Self-host it.
status: 🍑 Always yelling, never shuts up about animeI'm Jae Jin, better known online as Kadantte. I love designing and building software from the interface all the way down to the infrastructure it runs on.
My work lives at the intersection of web applications, DevOps, security engineering, automation, local LLMs, and GPU computing. I enjoy running my own services, understanding systems deeply, and making tools that are both practical and pleasant to use.
A good part of my work happens in private or self-hosted repositories, so this profile mixes public open-source projects with a short overview of ongoing builds.
Shiori is my in-development anime, manga and manhwa companion. She brings a little cozy Fuji-and-sakura magic to detection, approval, and tracking. 🌸
- Detect & suggest: identify a likely anime while allowing corrections with Choose Another Anime.
- Human-in-the-loop: only track after Approve & Track confirmation.
- Track intentionally: align playback-progress updates with the title you approved.
- Build carefully: focus on robust testing, privacy-conscious choices, and transparent behavior.
Shiori is an ongoing project. These describe planned and developing workflows, not a claim of public release.
| 💻 Build | 🛡️ Secure | 🏠 Run | 🤖 Explore |
|---|---|---|---|
| Full-stack apps, APIs, desktop apps, UX | Authentication, audits, MFA, defense in depth | Proxmox, TrueNAS, Docker, Forgejo CI/CD | Ollama, CUDA, PyTorch, OCR, embeddings |
| TypeScript, Python, Rust, PHP, C# | Rate limiting, RLS, secret hygiene | Linux, monitoring, networking, backups | Self-hosted LLMs, retrieval, computer vision |
🖥️ Peek inside my homelab
- Asura, my development workstation: Ryzen 9 7900 + GeForce RTX 3080 Ti, Windows 11 and Linux tooling.
- GPU lab: Threadripper PRO platform with dual RTX 5090 GPUs for local model experimentation and inference.
- Infrastructure: Proxmox, virtual machines, TrueNAS storage, Docker workloads, Forgejo, CI runners, reverse proxies and observability.
- AI & data: Ollama, PyTorch/CUDA, PostgreSQL, Redis, Neo4j and Qdrant, with a strong preference for local-first systems.
Details about access, internal networking, credentials, and security configurations stay private. 🔐
Selected ongoing work — some projects are private, experimental, or not yet released.
| Project | What I'm working on |
|---|---|
| 🔑 Kagishiro | A secure, self-hosted Ollama API gateway with API keys, quotas, streaming, observability and an admin dashboard. |
| 🌸 Shiori | An anime/manga/manhwa companion with title detection, Choose Another Anime, Approve & Track, and playback progress (in development). |
| 🧠 Advisio 2.0 | AI-powered application development spanning identity, email integrations, knowledge workflows and translation features. |
| 🛡️ Kleeja security | Research and proposed hardening for PHP administration: adaptive login protection, cross-IP abuse detection, MFA and passkeys. |
| 📡 Telematics & tracking | Integrations and protocol research around Teltonika devices, Traccar and custom data backends. |
| 🏗️ Home infrastructure | Automating and securing self-hosted apps, GPU inference, deployment pipelines and system monitoring. |
I care about secure defaults, real tests, maintainable architecture and shipping with intention.
Kurumi · C# Discord bot | Mikazuki · AniList desktop tracker | Candy Website · Personal web project
📚 I also explore anime/AI research collections, maintain forks, and contribute to other projects, including
MALClient.
DeepLearningAnimePapers and Suwayomi-Server are forks, not original projects authored by me.
The first two cards reflect badges from the previous profile; the remaining cards are engineering interests, not GitHub-awarded trophies.